IMPROVED METHODOLOGY FOR SQUEAK &

RATTLE ANALYSIS WITH ABAQUS AND

CORRELATION WITH TEST RESULTS

Eduard Caamaño, Inés Lama, Andreas Rousounelos, Jordi Viñas

IDIADA Automotive Technology, Spain

Abstract: : In recent years car manufacturers have been working intensively on new ways to
improve the quality of the interior trim since it is extremely important to the customer’s perception
of quality and can be a source of after-sale complaints. Consequently, the study of squeak and
rattle (S&R) has become one of the main concerns for car manufacturers. Simulation of S&R
phenomena is a complicated issue to reproduce virtually, because they are difficult to study using
methods based on eigenmodes due to the impossibility of using contacts in this type of FE model;
since modal theory is based on the hypothesis of linearity. Software providers and automotive
engineering companies have been working hard to provide tools and methodologies in order to be
able to analyse these phenomena in the initial development phases. In this context, in 2009
IDIADA developed a simulation protocol that took advantage of software already existing on the
market (Abaqus) by combining its capabilities to set up a methodology for detecting and analyzing
potential S&R problems. This paper presents the latest improvements and correlation of the
methodology of S&R detection. The improved methodology was applied to a complex structure (an
instrument panel) and the results were compared with measurements.
Keywords: NVH, Squeak, Rattle, Connectors, Postprocessing, Automotive Structures, Instrument
panel, Trimmings Development

1. Introduction

In recent years, with the improvement of the acoustic quality of vehicles and engines, other types
of noises that normally were hidden by the usual noise of the engine and the rolling noise are now
perceived by the occupants. Some of the most annoying ones are the small noises coming from the
squeak and rattle of the interior trimmings, which in general have become an indicator of the
quality and durability of the product.
Squeak phenomenon is caused by the relative movement between two components that slip and
stick alternatively. The elastic deformation of the surfaces in contact accumulates energy and this
is released when the static friction overcomes the kinetic friction producing then an audible noise.
The generation of this type of noise occurs at low frequencies (under 200 Hz) they are normally
induced by energy inputs coming from the suspension system; however the release of the elastic
energy produces the vibration of the adjacent surfaces causing contact audible noises in the range
of 200 to 10000 Hz (Kavarana, 1999).

2011 SIMULIA Customer Conference 1

Rattle noise is induced by impacts between parts; it is generated by the perpendicular relative
movement between two parts, and is usually due to erroneous assembly tolerances or to lack of
stiffness. The impacts will only generate noise if the adjacent surfaces to the impact point are
capable of radiating audible sound-power levels (Kavarana, 1999).
Traditionally, automotive industry has been using physical testing to detect this kind of defects,
but lately some efforts have been put in developing software and methodologies to simulate these
phenomena. In 2008 a relatively simple methodology was developed based in ABAQUS for the
simulation of the rattle event. This methodology was based mainly in the use of the Connector
element of ABAQUS which was used as virtual sensor for the detection of the contact.
In that methodology the Connectors were created automatically in all the areas that can experiment
contact events by defining a sphere of influence or a radius suitable for the model in relationship
to the clearances of the structure.
During the simulation all basic and required information for the Squeak & Rattle analysis are
stored in the Connector element. The connector section is defined as BUSHING and in that way, a
connection between two nodes is created such as it allows independent behaviour in three local
Cartesian directions that follow the system at both nodes a and b. Three amplitudes expressed in
three local Cartesian directions can be obtained:

• Displacement amplitudes expressed in local directions 1 and 2, which are tangential to

the connector.

• Displacement amplitude expressed in local direction 3, which is coincident with the axis
of the connector itself.
With the normal NVH model plus the virtual sensors any type of vibratory simulation can be
launched (direct response, modal superposition,…). In the time domain the virtual sensors
wouldn’t be necessary because it is possible to define contacts between pieces, but in the
frequency domain, they are the only way to detect where a contact will probably arise. The
connectors allow tracking the steady-state amplitude of the movement relative to the two parts
which the sensor is connected for each value of frequency in the range of study. In other words, it
will be possible to study the resulting penetration between parts for each frequency. This value of
penetration is the key to evaluate the chance of squeak or rattle.
The last step is the post processing. As mentioned before, the virtual sensors provide the amplitude
of movement between two parts. Nevertheless, this value by itself alone has no valuable
information for the final objective. The chance of squeak or rattle is not determined by the value of
amplitude alone. The value of penetration must be calculated to do so.
In the methodology developed in 2008 the chance of rattle was related to the value of the
calculated penetration. The chance of rattle, was automatically calculated using scripting and
stored in a results file for its graphical post processing. Rattle was represented in the first version
of the methodology as vectors being their magnitude the value of penetration between parts. This
kind of visualization allows a rapid detection of problematic areas and an easy understanding of
the phenomena by means of the analysis of the movements. For a better visualization it is possible
to sweep the frequency within the whole range of study and follow the appearance of the rattle

2 2011 SIMULIA Customer Conference

depending on the frequency. Another possibility is to analyze the most critical frequencies using a
2D plot. Using scripting an ASCII file is created with a sum up of the rattle issues that appear in
the whole frequency range of study.
The script scans all connectors’ responses and searches the frequency of the maximum penetration
detected for all the connectors and saves the frequency of appearance and the value of amplitude.
Such file can be visualized in a variety of spreadsheets in order to analyze the frequency ranges
where the phenomenon appears.

FE Model

FE Model

Connector local directions

Virtual sensor automatic generation Gap between parts

Analysis

Modal analysis
Relative amplitude displacement
Modal superposition

Results analysis

Results generation

Analysis of problematic frequencies Analysis of problematic areas

Maximum penetration (mm)

20 40 60 80 100
Frequency (Hz)

2D Plot: maximum interference (f) 3D Plot: vectorial visualization of the interference

Squeak and rattle detection

Figure 1. Workflow of the methodology

2011 SIMULIA Customer Conference 3

The main workflow of the methodology is shown in Figure 1.
With the work done since 2008, some improvements have been introduced in the methodology in
order to validate it for complex structures:
- the accuracy of the interference detection has been increased
- the automatic creation of the connectors has been sped up and their location is now more realistic
- the criterion of rattle detection has been also improved
Finally, as it will be explained in the following paragraphs, a complex automotive structure has
been tested within the methodology in order to establish the modelization rules required for an
acceptable degree of correlation of the rattle. In that process, the weak points of the methodology
for its usage in normal development projects were also pointed out.

2. Improving the methodology

The basic methodology was developed in 2008 taking as sample a simple piece (a wheel trim),
however in order to assure its robustness more complicated components were also tried. The
instrument panel of the new SEAT Ibiza was taken as sample and was used for validating the
methodology and correlating the rattle results.

Figure 2. Instrument panel used for the study, courtesy of SEAT.

2.1 Rattle detection criteria

In the first version of the method it was assumed that the rattle ratio was equivalent to the
interference measured by the Connectors, however the following aspects should be taken into
account:
- at low frequencies the amplitude of movements is normally higher than at high frequencies and,
therefore, the calculated interference will also be bigger. However, the low frequency movements
have low associated velocities and, therefore, the kinetic energy in the impact can also be low (low
dissipation of energy through the rattling event).
- at high frequencies the displacements are normally lower, but the velocity in the instant of the
impact can be high.
For the previous reasons, it was considered that the evaluation of the kinetic energy at the impact
time and its usage as scale factor for the interferences could offer a better approach as rattle

4 2011 SIMULIA Customer Conference

criterion; therefore, the rattle ratio is calculated as the product of the interference detected and the
kinetic energy at the impact time. All detected rattle ratios are afterwards scaled in reference to the
highest rattle detected. It must be noticed that the rattle ratio is a relative value, it has no physical
meaning; it shows only a bigger chance of having a rattle in one zone than in another according to
the interference degree and the kinetic energy of the impact.
The interference in each connector is calculated as the difference between the longitudinal
movement of the connector and the initial gap between the pieces. In point 2.2 it is explained why
in the new version the gap is directly the elongation measured by the connector, and in that way
the procedure has also been sped up.

2.2 Clearance analysis

In the previous version of the method, connector elements were created directly from mid-
surfaces. This way of procedure implied that thicknesses of the different parts had to be taken into
account in order to calculate the real gap between parts.

Figure 3. Evaluation of the gap in parallel elements or in elements with angle.

When parts are parallel, the calculation of the gap value is straight forward; half of the thickness of
each part is subtracted from the connector length. However, when parts are not parallel one to each
other, the relative angle between parts plays an important roll in the gap value. This fact makes the
calculation of the gap value more tedious.
It has been seen that the correct calculation of gap value is essential since the final value of rattle
is directly dependant on it. If the gap value is incorrect, the results for the rattle will also be
incorrect.
In this sense, after studying different proposals, it has been accepted that the best way to procedure
is to volumize the mid-surface shells so that we work with a model which is more representative
of the real geometry. Working this way, the connector lengths are directly the gap between parts
since the relative angle is implicitly taken into account.

2011 SIMULIA Customer Conference 5

 Figure 4. Effect of the volumize tool.

It is also interesting to note that with the new technique we make the routine more precise since
more connectors are created. Besides, since the gap value is directly the connector length and there
is no need to subtract any thickness from it, the time execution of the routines has been improved
significantly.
The main disadvantage of generating the connectors using the volumized mid-surface shells is that
more connectors than the ones strictly needed are created. For this reason, the routines have been
adapted in order to delete those connectors that either have no physical meaning or give no
additional information for squeak and rattle purposes.

3. Improving the correlation

The detection of the rattling event is based on an accurate simulation of the frequency response
functions (FRFs) of the areas susceptible of having a contact event. For that reason, a big effort
has been put in improving the correlation not only in terms of the eigenmodes but also in terms of
frequency responses.
The improvement of the correlation has been carried out in two steps. In the first one the
eigenmodes of the whole instrument panel as a global model and also by different areas were
studied and compared with the experimental ones. Table 1 shows the Modal Assurance Criterion
(MAC) values of different modal pairs. The MAC values have been calculated for the whole
structure as well as for specific areas in an effort to localise the areas of higher discrepancies. All
MAC value calculations have been performed on the same test and simulation data.

6 2011 SIMULIA Customer Conference

 Test Sim. Global MAC MAC MAC MAC
Freq. Freq. MAC without without only only
(Hz) (Hz) HVAC glovebox outer MQT
panel

1 28.8 28.1 0.77 0.83 0.83 0.91 0.69

2 33.1 34.2 0.51 0.54 0.53 0.61 0.71

3 40.7 36.6 0.70 0.74 0.71 0.77 0.67

4 43.3 40.4 0.57 0.63 0.70 0.80 0.63

5 93.9 82.1 0.53 0.59 0.56 0.61 0.52

Table 1. Global and partial MAC values.

The correlation of a model with test results of a complex part, such as an instrument panel,
possesses several challenges because of the complicated connections of different components.
Several modifications were tested and implemented in order to improve the correlation of the
instrument panel under study compared to previous published result (Lama, et. al., 2009). The
main modifications are the addition of more geometrical details, including the masses of the
radiators in the HVAC system, as well as the use of more appropriate values for the stiffness of
different connection points.
Three connection points were found to influence the results the most and were studied more
intensively. These points are the two connections of the HVAC ducts leading to the two side
diffusers and the connection of the HVAC duct leading to the central diffuser. The values of the
stiffness of these connection points were determined through an optimisation procedure which
aimed to minimise the difference between measured and simulated FRFs for all the measured
points.
In the following pictures some examples of the degree of correlation of the FRFs are shown. The
modifications introduced in the model led to better results not only in the MAC values but also in
the FRFs comparison.

2011 SIMULIA Customer Conference 7

 Figure 5. Testing layout of the instrument panel.

(a) HVAC X-direction (b) Dashboard right Y-direction

(c) Cross car beam Z-direction (d) Glove box Z-direction

Figure 6. Comparison of measured FRF (in red) and simulated (in blue) for four different points

8 2011 SIMULIA Customer Conference

4. Rattle results

A listening test was performed on the real structure and the dominant areas of S&R generation
were identified for different frequencies. A rattle simulation was also performed for the same
excitation as in the listening test. The simulation was able to predict the rattle areas with a
reasonable degree of correlation. Figure 7 shows However, there were more rattle areas detected
by the simulation compared to listening test. Some of the reasons for this discrepancy, regarding
the simulation procedure, could be small differences between the mesh of the FE model and the
real structure, especially regarding clearances, the generation of some misplaced rattle sensors
during the automatic generation procedure and the poor correlations at some part of the model.

Figure 7. Rattle detected in the connections of HVAC tubes in the frequency range 41Hz-46Hz

2011 SIMULIA Customer Conference 9

 Figure 8. Rattle detected in the glove box in the frequency range 71Hz-75Hz

5. Conclusions and future work

In the present work an additional step has been done in the area of simulating S&R. Some
improvements have been introduced in the methodology and also in the vibrational correlation in
order to obtain better results. One of the main conclusions is that a more realistic representation of
the geometry helps to improve the accuracy of the rattle responses. This leads to the conclusion
that, when an S&R analysis is planned, accurate meshes should be obtained from the first stages of
the project. This can be achieved by using advanced software capabilities (like continuum shells)
and automatic meshing technics. An accurate mesh will avoid re-meshing work at a later stage of
the project and will ensure correct clearance representation for a correct S&R analysis.
The simulation procedure for rattle detection presented here was able to predict some of the
dominant areas where rattle generation was observed during a listening test. In the future, the aim
is to use the rattle simulation in more applications in order to fine tune several aspects such as the
automatic generation of rattle sensors and the impact detection calculation procedure. Moreover, a
more precise procedure for rattle detection during testing is required to give useful feedback for
the results obtained through simulation.

6. References

1. El-Essawi, M.., Lin, J. Z, Sobek, G., Naganarayana, B.P. and Shankar, S., “Analytical
Predictions and Correlation With Physical Tests for Potential Buzz, Squeak, and Rattle
Regions in a Cockpit Assembly”, Society of American Engineers International, Detroit,
Michigan, March 8-11,2004

10 2011 SIMULIA Customer Conference

2. Kavarana, F. and Rediers, B., “Squeak and Rattle – State of the Art and Beyond”, SAE paper
1999-01-1728, Noise and Vibration Conference, Traverse City, Michigan, May 17-20, 1999.
3. Naganarayana, B.P. and Shankar, S., Battachar, V. S., Brines, R.S. and Rao, s. , “N-hance:
Software for identification of critical BSR locations in automotive assemblies using finite
element models”, SAE paper 03NVC-283, Noise & Vibration Conference, Traverse City,
Michigan, May 5-8, 2003.
4. Lama, I., Viñas, J., Blecon, Y., Montané, X., “Analysis of S&R phenomena through
simulation in ABAQUS”, SCC2009

7. Acknowledgements

The authors gratefully acknowledge SEAT for giving all the necessary information for the present
work.
They also acknowledge Pau Cruz for his useful ideas for the development of the methodology.

2011 SIMULIA Customer Conference 11